From owner-freebsd-hardware Wed Aug 28 23:35:03 1996 Return-Path: owner-hardware Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id XAA19272 for hardware-outgoing; Wed, 28 Aug 1996 23:35:03 -0700 (PDT) Received: from tide19.microsoft.com (tide19.microsoft.com [131.107.3.29]) by freefall.freebsd.org (8.7.5/8.7.3) with SMTP id XAA19254 for ; Wed, 28 Aug 1996 23:34:56 -0700 (PDT) Received: by tide19.microsoft.com with SMTP (Microsoft Exchange Server Internet Mail Connector Version 4.0.993.5) id <01BB9539.77D67B00@tide19.microsoft.com>; Wed, 28 Aug 1996 23:34:33 -0700 Message-ID: From: Thomas Pfenning To: "'Rodney W. Grimes'" Cc: "'richard@pegasus.com'" , "'freebsd-hardware@FreeBSD.org'" Subject: RE: DAT or removable device? Date: Wed, 28 Aug 1996 23:33:47 -0700 X-Mailer: Microsoft Exchange Server Internet Mail Connector Version 4.0.993.5 Encoding: 43 TEXT Sender: owner-hardware@FreeBSD.org X-Loop: FreeBSD.org Precedence: bulk That is an interesting measuring method. The August issue of Byte published a comparison of Ultra SCSI adapters and they had the AHA-2940 Ultra wide and a Symbios 53C875 based board on the test. They tested it with 8 drives and at least the 53C875 matches or beats the Adaptec in all tests. The Adaptec actually came in last in accumulated bandwidth. These tests where performed under Novell Netware, however, the accumulated > 36MB/s with 8 disks on a single UltraSCSI card seem to indicate that this is pretty close to the saturation of the SCSI bus. Of course, the Adaptec came in first with the highest price:-) Cheers Thomas >---------- >From: Rodney W. Grimes[SMTP:rgrimes@GndRsh.aac.dev.com] >Sent: Wednesday, August 28, 1996 9:33 PM >To: richard@pegasus.com >Cc: freebsd-hardware@FreeBSD.org >Subject: Re: DAT or removable device? > >> } * Supposedly, the NCR/Symbios-based cards don't perform well under heavy >> } loads, such as when you have three or more disks being accessed >> } *simultaneously*. This isn't an issue for most people. >> >> This appears to be hearsay. No numbers have been presented. > >This is not hearsay. I can't give you numbers that mean a whole lot, >they are from running a PCI bus analyzer. The busier a NCR 53Cxxx >gets the more PCI bandwidth, and host memory bandwidth gets eaten >up. And I am not talking about the data transfer bandwidth (though >that is the majority of the PCI traffic), but the microcode fetch >bandwidth. > >I do not recommend running lots of drives or lots of 53C8xx's in a machine >for that very reason, UNLESS lower cost is an objective and performance can >suffer by reducing the cost. > >-- >Rod Grimes rgrimes@gndrsh.aac.dev.com >Accurate Automation Company Reliable computers for FreeBSD >